The first production crawl of noverva.com found fifty pages missing canonical URLs, three false positives in our own rule engine, and a bug that had been reporting build-time health for a year.